A Supervised Learning Estimator And NLP Framework For Detection of Cyber Fraud in Promotional Content

The dissemination of intentionally deceptive content disguised as legitimate journalism is a global issue undermining information accuracy and integrity. This phenomenon significantly influences public opinion, decision-making, and voting patterns. Most false news originates on social media platforms like Facebook and Twitter, later infiltrating mainstream media outlets such as television and radio. These false stories often share linguistic traits, including excessive use of unsubstantiated hyperbole and non-attributed quotes. This paper presents the findings of a study on false news detection, focusing on a novel classifier developed using Textblob, Natural Language Toolkit (NLTK), and SciPy. The system employs quoted attribution as a key feature within a Bayesian machine learning framework to estimate the likelihood of an article being false. The methodology achieved a precision rate of 63.333% in identifying false articles containing quotes. This innovative process, termed "influence mining," is introduced as a potential tool for detecting false news and propaganda. The study details the research process, technical analysis, linguistic features, and classifier performance. It concludes with insights into the evolution of the current system into a comprehensive influence mining framework capable of addressing broader disinformation challenges.